Download the interview, or listen here:
SalesPredict – Predicting Revenue In Your Future
Think that future prediction is a field taken from science fiction? Well, it is time to challenge your thinking!
For Dr. Kira Radinsky – a co-founder and the ‘brain’ behind Israeli startup SalesPredict, the future can well be predicted, if you only look for answers deep enough in the data.
“In my early days in academia I was involved in a research on prostate cancer, in which we were trying to predict which nutritional substances affect the likelihood of developing this disease” she explains in an interview for Startup Camel. “And as part of the research we had to count manually thousands of cells – a very cumbersome task. So that was when I built my first computer program that would count cells for me”.
Several years later, already a PhD researcher, Radinsky decided to apply her machine learning knowledge to another, surprising field: “I wanted to see whether we can collect data from all past news coverages – as early as year 1500 – and use it to predict future events, the way we predicted the likelihood of developing cancer based on past nutrition. We programmed our computers such that they would read all these past newspapers like humans do. And we combined all that data and thought – what can we learn from it about the future?”
How good were the results? Well, Radinsky’s predictions have been accurate enough in at least several events, and have drawn enough traction from investors, that she eventually decided her to put her academic research aside and start a company.
Today, SalesPredict leverages machine learning technology and economic modelling to help other companies increase sales by predicting where their business is going. “We help B2B companies, specifically salespeople, examine potential business interactions by looking at the probabilities that these interactions will succeed. Based on our results companies can choose their optimal business path or sales strategy”, explains Radinsky.
Count on your Machine Learning, not your gut feeling
Predicting the future is one thing, and even if you’re skeptical – SalesPredict’s results will often prove you wrong. Yet, as a startup that focuses on selling its predictions to salespeople, the real challenge is not in the predictions themselves – but rather in how to convince salespeople that their key insights are in the data, rather than in their own gut feeling. “To overcome this challenge, we realized two things”, explains Radinsky: “first, we had to do a better job in explaining our modelling to salespeople. Second, different explanations work for different people, and we had to find a way that would enable us to identify the right explanation for each specific salesperson”.
With these understandings, Radinsky and her team went on to build another model – Emotional Artificial Intelligence. The model helps SalesPredict learn the personality traits of each salesperson (who is a potential customer for SalesPredict), then tailor the explanation most likely to convince that salesperson that SalesPredict’s prediction should be taken seriously. “In a way, even emotions (to a certain level) can be modelled. Some people say they like working with gut feeling rather than with data. But in reality, gut feeling is a set of past experiences that help the salesperson estimate the outcome of potential future actions. This is exactly how machine learning works – it looks at historical statistics and makes predictions based on that data”.
So what does the future hold for SalesPredict?
Although predicting business outcomes is the bread and butter for SalesPredict, Radinsky has a bigger goal than just maximizing sales: “Much of the economic study today is based on theory only, with no real data-life data to support and validate the underlying assumptions. But with all that data we have at SalesPredict about what works and what doesn’t work for companies, we can move forward science as well. The more we investigate companies and businesses, the better understanding we have of how economy works – not just in theory, but in practice. So in the future, we want to complete the missing piece for economic modelling – the piece of real-life data”.
After all, even if you cannot test your economic model tomorrow morning, it might as well have been tested by reality. Just look where the answers are – in past data.